Probabilistic State Parameter–Based Liquefaction Triggering Model

Abstract

Soil liquefaction is a central concern in geotechnical engineering because of its potential to damage infrastructure. Current liquefaction triggering practice relies on semiempirical procedures calibrated from earthquake case histories and surface manifestation labels, which can be biased and unevenly distributed across seismologic, geologic, and geomorphic settings. This paper develops an alternative probabilistic triggering framework anchored in laboratory testing, formulated in the state parameter space, and evaluated using independent piezocone penetration test (CPTu) case-history databases. Cyclic resistance is calibrated across 20 sands using cyclic and monotonic triaxial test data through a hierarchical Bayesian regression that separates within-material dispersion from between-material heterogeneity and propagates specimen-level uncertainty. A CPTu-based procedure is implemented to compute distributions of factor of safety F S and probability of liquefaction P L via Monte Carlo uncertainty propagation, jointly accounting for capacity, demand, and state parameter estimation. Receiver operating characteristic–based evaluation indicates discrimination comparable to, and in some cases exceeding, widely used CPTu procedures. A single-material calibration example shows that incorporating site-representative laboratory data can reduce cyclic resistance uncertainty by an average of 80%. This supports a workflow in which population-level predictions are useful for screening, while material-specific calibration is recommended for critical projects.